Fog Creek Software
Discussion Board




reverse engineering

I have been asked to reverse engineer a system, that is look at the output, and find the algorithm generating that output.
Is there any general procedures for doing this?How far has reverse engineering been successful?

Anonymous
Monday, August 04, 2003

what kind of system is it?

anon
Monday, August 04, 2003

data mining one

Anonymous
Monday, August 04, 2003

tip: Take the input into account as well
;)

Geert-Jan Thomas
Monday, August 04, 2003

9 out of 10 times its easier just to work out your own algorithm for doing things.

FullNameRequired
Monday, August 04, 2003

With data mining it can be pretty hard to work it out yourself - for example it's not going to be easy to work out how to induce decision trees with C5 by yourself.  Say you're trying to reverse engineer SAS or SPSS to work this out, that's a very difficult task. 

You could try checking out the source for Weka if you haven't already, it might be easier to go for something that is 'second best' and taking it from there from a position of real understanding.  Source for C4.5 is in the public domain in C as well if I remember rightly, and I believe Weka would yield the goods for association rules/sequences, and k-means clustering etc. 

Konrad
Monday, August 04, 2003

neural networks

Ross
Monday, August 04, 2003

Neural Networks aren't the be-all-end-all. Sometimes a simple decision tree is the right way to go, and less processor intensive, and just as accurate.

You also have to have a good sampling of data to tune your neural network, and know which data points are important and which ones to ignore.... otherwise it gets really processor intensive as you try to data mine every particle in the known universe to determine the feasability of the existance of God.

Wouldn't a simple "faith / no faith" question be easier? lol.

www.marktaw.com
Monday, August 04, 2003

*  Recent Topics

*  Fog Creek Home